Machine learning in trading: theory, models, practice and algo-trading - page 2735
You are missing trading opportunities:
- Free trading apps
- Over 8,000 signals for copying
- Economic news for exploring financial markets
Registration
Log in
You agree to website policy and terms of use
If you do not have an account, please register
If there is no activation in the leaves, it means that the current sample does not correspond to the one area on which the model was trained.....
If the current sample does not match, it means that it is different, and this is the fact of its change.
What is the problem with substituting your own data into my script?
I don't understand what it should do in the end. We need it to determine the optimal length of a section from the whole sample, will it do that?
It is even desirable to break the sample into sections for consistent application and training.
To make the participants of the discussion understand the ideas better, as practice shows, everyone here has very different representations, even at the initial concepts. The closeness of models of objects to the objects themselves depends on the representations.
He's an idiot, don't even try to explain anything.
I don't understand what he's supposed to do at the end of the job. We need it to determine the optimal length of the section from the whole sample, will it do that?
It is even desirable to divide the sample into sections for consistent application and training.
I don't understand what you want to do and how you want to do it, how to determine what you want to do.
Yeah I don't understand what you want to do and how you want to do it either, how to define what
Determine the minimum window or sample size of a meaningful target. I don't know how. Maybe by brute force, but that's expensive. It's hard to think of one off the top of my head. It seems to be necessary that all the desired features should be in the sample and enough once, may be repeated, but that all at least once. I don't know how to determine this without knowing the features.
Determine the minimum window or sample size of a significant target. I don't know how. Maybe by brute force, but it's costly. It's hard to think of one. It seems to be necessary that all the desired features should be in the sample and enough once, may be repeated, but that all at least once. I don't know how to determine this without knowing the features.
The goal is clear...
But honestly, I don't see any profound idea in it.
The purpose is clear.
But honestly, I can't for the life of me see any profound idea in it.
What is the purpose of "representations"? What is the purpose of "representations"?
If it is philosophical, there is no question.
But in financial markets there are only two purposes: to predict the value and to predict the direction (sign).
If the "representation" is for this, then what influence, how related, what predictive power do all these "representations" have for the purposes outlined above?
The method of predictions should be further clarified. If they are made by guessing on the entrails of animals, then a source of animals is needed - an animal farm, for example. If mathematics is used for predictions, then the appropriate mathematical apparatus is needed, based on concepts that are usually introduced on the basis of ideas about the subject area.
It is necessary to specify the method of predictions. If they are made by means of guessing on animal entrails, then we need a source of animals - an animal farm, for example. If mathematics is used for predictions, then the appropriate mathematical apparatus is needed, based on concepts that are usually introduced on the basis of ideas about the subject area.
Firstly, the new-speak that obscures the point.
Second. In MO there is a concept, probably "representation" - predictor and a bunch of synonyms for it.
I pause to reiterate my understanding: there are predictors that are relevant to the target with different levels of strength, and there is rubbish. What is what can be learnt only from the target-predictor bundle. In my models, the initial list of predictors is about 180 pieces. Prediction is performed on no more than 10, and the rest is rubbish, and the names of predictors included in 10 change as the window moves.
Hence my question about "representations". Why do this outside the "predictor-target" linkage and without a measure of attribution to rubbish (animal entrails).
Firstly, newspeak that obscures the point.
Secondly. In MO there is a concept, probably "representation" - predictor and a bunch of its synonyms.
I pause to reiterate my understanding: there are predictors that are relevant to the target with different levels of strength, and there is rubbish. What is what can be learnt only from the target-predictor bundle. In my models, the initial list of predictors is about 180 pieces. Prediction is carried out for no more than 10, and the rest is rubbish, and the names of predictors included in 10 change as the window moves.
Hence my question about "representations". Why do it outside the "predictor-target" linkage and without a measure of attribution to rubbish (animal entrails).
No new-speak, of course. Representation as an image of an object in the mind is a long-used term. Representing an object as a string of numbers is a recent and deeply secondary concept. The list of predictors itself is built on the basis of our personal perception of the object and therefore it will never coincide in two different "machinists". This, by the way, is one of the reasons why a meaningful discussion of specifics is hardly possible in this thread.
Please tell me, or at least give me a hint (wink your eye) - have you managed to build a working earning Expert Advisor on a neural network?